Retroactive public goods funding (RetroPGF, later renamed Retro Funding) is a grant-making method that reverses the usual order of operations: instead of a committee predicting which projects will be valuable and funding them in advance, it waits until work has been done, judges what actually proved useful, and pays for that. The idea was set out by Optimism in July 2021 – with a section contributed by Vitalik Buterin – and rests on a single claim: “it’s easier to agree on what was useful than what will be useful.”
The appeal is obvious to anyone who has run a grants program. Prospective grants are bets on people and pitches, and most of the information that would settle the bet arrives after the money is gone. Retroactive funding waits for that information. Between 2021 and 2025, Optimism spent tens of millions of OP tokens testing whether the thesis survives contact with a real treasury and real voters. This page covers the mechanism, the rounds, what Optimism itself concluded went wrong, the January 2026 pause, and the handful of other ecosystems that have tried it.
The original proposal has two halves, and the second is routinely forgotten. The first half is the results oracle: “A DAO, which we can call ‘the Results Oracle’, funds public good projects… retroactively, rewarding projects that it recognizes as having already provided value.” That is the part everybody implemented.
The second half is what was supposed to make it more than a well-timed grants committee. Projects issue tokens, and because everyone knows a results oracle will eventually pay out for demonstrated impact, those tokens become – in the post’s own words – “essentially creating a prediction market for what the results oracle will fund.” Speculators who believe a project will later be judged valuable buy its tokens early, which funds the work now. The oracle sets a floor: “if it allocates a reward of $X, and the project has a total supply of N tokens, then it publishes an open order to buy up to the entire supply of those tokens at a price of $X/N per token.” Retroactive judgement, prospective capital. The oracle never has to predict anything; the market does the predicting and takes the risk.
Optimism folded the ambition into its Working Constitution (April 2022) as a stated purpose: “The primary function of the Collective is to minimize the discrepancy between collective impact and individual profit” – the slogan impact = profit. In practice the funded-project-token market never materialised at scale, and Retro Funding ran as the oracle alone: a periodic vote distributing a pot to projects that had already shipped.
Optimism ran six numbered rounds. The scale grew roughly thirty-fold, and the mechanism changed almost every time – each change an attempt to fix what the previous round exposed.
| Round | When | Pot | Voters | Mechanism |
| RetroPGF 1 | Q4 2021 | $1M (pre-OP-token) | 24, picked by the Foundation | Ballot via quadraticvote.co; 58 of 76 nominees paid |
| RetroPGF 2 | March 2023 | 10M OP | 69 of 71 voted | Mean; all 195 nominated projects paid |
| RetroPGF 3 | Nov–Dec 2023 | 30M OP | ~145–146 | Median with a quorum of 17; 1,594 applied → 501 paid |
| Retro Funding 4 | July 2024 | 10M OP | 108 of 133 voted | Metrics-based – voters weighted metrics, not projects |
| Retro Funding 5 | Oct 2024 | 8M OP | 111 (84 Citizens + 27 guests) | Category subgroups |
| Retro Funding 6 | Nov 2024 | 2.4M OP | 78 of 102 Citizens + 60 of 76 guests | Guest votes at 1:1 |
In 2025 the round structure was replaced by continuously-measured missions: Dev Tooling and Onchain Builders, each up to 8M OP, evaluated monthly against onchain metrics rather than by a vote on projects. Optimism described the programme at that point as backed by “850M OP dedicated to Retro Funding” – one of the largest sums any ecosystem has earmarked for public goods.
The trajectory is worth naming plainly: each redesign removed human judgement. Round 3’s voters judged projects directly and drowned. Round 4 had them judge metrics instead. The 2025 missions had an algorithm judge continuously. The mechanism kept solving its voter problem by giving voters less to do.
The most useful critique of Retro Funding is not external. Optimism’s RetroPGF 3 retrospective is unusually candid, and every finding below is the Collective’s own.
Vitalik’s own review of Round 1 had flagged the same failure four years earlier: winners were “more well-known projects”, plus a friendship bias and a monoculture in which essentially every winner was a technology project. The design never shook it. Academic work has since shown that all four of the voting mechanisms Optimism used are not strategyproof, and that the Round 3/4 quorum rule induces formal failures of monotonicity and participation.
On 8 January 2026, in the post announcing Season 9, Optimism stated: “The Retro Funding program will not run for at least the next 12 months.” The rationale given: “As Optimism enters a new phase focused on execution, scale, and enterprise adoption, the Retro Funding program will be paused to reevaluate how it fits into Optimism’s long term strategy.” The governance-side Guide to Season 9 (7 January 2026) is blunter: “The Foundation will not propose any Retro Funding Missions in Season 9 or 10.”
The reserved tokens are in play. Per the Season 9 guide, “A proposal to re-allocate all or a portion of the tokens reserved for Retro Funding (~775M OP), and/or other uses, will be put forward by the Foundation during Season 9.” (The blog hedges the same sentence to “may be put forward… in the coming months”.) At the time of writing this wiki has not verified whether that reallocation vote was held or how it resolved – Season 9 ran to 3 June 2026.
Language matters here, so: Optimism has paused Retro Funding, not cancelled it, and says “Public goods remain core to Optimism’s vision.” The 12-month floor runs to at least January 2027. But the plain reading is that the largest retroactive-funding experiment in crypto – 850M OP earmarked, roughly 20M OP paid out in 2024 alone – is stopped, its remaining ~775M OP is a candidate for reallocation toward enterprise adoption, and the reason given is not that the mechanism failed but that the organisation’s priorities moved. Governance in Season 8 had already removed the last incentive to participate: “There will be no financial rewards or incentives for being a Citizen or voting in Season 8 and beyond.”
Retro funding travelled, though most copies are smaller and several “retro” programmes are retroactive in name only. Three distinctions are worth holding.
Genuinely retroactive – impact judged after the fact:
Retroactive target, quadratic mechanism – closer to quadratic funding pointed backwards: Arbitrum’s Citizens Retrofunding Round 1 (100k ARB, late 2023), and Gitcoin’s own Citizens Retroactive Funding ($153k in GTC). Gitcoin also supplied much of the tooling: WEST built a Round 3 frontend for Optimism and open-sourced it as EasyRetroPGF, the stack Celo and others reused.
Frequently miscategorised: Protocol Guild is continuous vesting rather than a round, but weights shares retroactively – “The earlier a contributor’s start_date, the higher their overall Split share, rewarding them retroactively for historical contributions.” Octant is often grouped with RetroPGF but its allocation code is quadratic funding over staking yield – prospective matching, closer to Gitcoin Grants. And Avalanche’s Retro9000 (up to $40M) is retroactive in disbursement – you build first, then get paid – without the impact-assessment vote that defines the model.
The 2021 thesis is probably still true: hindsight really is cheaper than foresight, and it is easier to agree on what was useful than what will be. What five years of rounds demonstrated is that this is not the binding constraint. Knowing the past is easy; agreeing on it at scale, under a deadline, with money attached, by people who are entangled with the projects they are scoring – that is the hard part, and retroactivity does nothing about it. The judgement problem was never in the tense.
The other lesson is structural. The half of the design that made retro funding elegant – project tokens as a prediction market that front-runs the oracle, so capital reaches builders before the reward lands – is the half nobody built. Without it, retroactive funding is a grants committee with better information and worse timing: the work still has to be financed somehow while it happens, and “ship first and hope a committee notices” is a funding model only the already-funded can afford. That, rather than any voting-mechanism defect, may be the deepest reason the winners kept looking like the projects that were already well-known.
Caper does not run a grants programme, and nothing in it is a retro-funding mechanism. It shares one aim with Retro Funding, that the record of what a member did should decide what they get, and applies it to a much narrower thing.
In a caper, each ballot a member casts earns them a non-transferable vote token, and nothing else earns one: trades and transfers never create it. A member's vote weight comes from that record and the tokens they hold together, so capital still counts, but a large bag alone can't capture control, and nobody can buy the record off a member who already holds it. The same weight decides what a member takes when they leave: a member who has voted can exit any time for their fair share of the treasury, which is Caper's standing form of the rage quit.
Set against Retro Funding's failure modes, the difference is who judges. Optimism's own retrospectives, covered above, found its badgeholders overloaded, drawn to projects they already knew and sometimes conflicted. A caper has no badgeholders, no committee and no impact assessment, because it records one thing mechanically: whether a member voted. That says nothing about whether they voted well, and it is no substitute for funding public goods; what it does reward, it rewards without anyone's judgement.